Big tech companies spend enormous sums of money all the time.
They buy back stock, pay dividends, make acquisitions, and write checks that would make a defense contractor blush. Most of that spending is legible to investors. You can model it, forecast it, and plug it into a spreadsheet.
But there is a specific kind of capital spending that breaks conventional financial modeling, and that is the kind where a company is building infrastructure for demand that doesn’t exist yet.
That is exactly what Microsoft (MSFT) is doing right now, and it’s why a new research note from Morgan Stanley should stop any MSFT investor in their tracks.
The question isn’t whether Microsoft is spending big. It’s whether Wall Street is giving it credit for what that spending will eventually produce.
Why Azure revenue estimates look too conservative
Morgan Stanley analysts Keith Weiss and Josh Baer published a report on May 27, 2026, framing Microsoft’s AI infrastructure expansion through a metric they call “revenue per megawatt” (MW) of datacenter capacity.
What struck me when I worked through their numbers is how clearly the math reveals a gap between what Microsoft is building and what analysts are currently forecasting as revenue.
The firm estimates Microsoft’s cloud ecosystem currently generates roughly $20-30 million in annualized revenue per MW of datacenter capacity, with that figure declining toward the high-teens by fiscal year 2028, according to Morgan Stanley.
More Tech Stocks
- Morgan Stanley sets jaw-dropping Micron price target after event
- Nvidia’s China chip problem isn’t what most investors think
- Quantum Computing makes $110 million move nobody saw coming
That sounds like a bad thing. It isn’t.
The firm argues the decline reflects one reality: Microsoft is deploying AI-specific datacenter capacity well ahead of the revenue that capacity will eventually generate. In other words, the electricity is already flowing. The money just hasn’t caught up yet.
Microsoft’s installed datacenter footprint is projected to scale from roughly 5 gigawatts (GW) in fiscal year 2024 to approximately 20GW by fiscal year 2028, based on Morgan Stanley estimates.
Management confirmed on the fiscal third-quarter 2026 (F3Q26) earnings call that the company “added another gigawatt of capacity this quarter,” according to the report.
Related: Bank of America resets Microsoft stock forecast after earnings
The $190 billion bet that changes the math
You have to understand the scale of what Microsoft is committing to before the revenue upside makes sense.
Azure revenue surged 40% in fiscal Q3 2026, beating analyst expectations for 38.8% growth, according to CNBC. Total revenue came in at $82.9 billion, up 18% year over year, which Microsoft confirmed beat consensus estimates of $81.4 billion.
Microsoft also disclosed that capital expenditures for fiscal year 2026 are now projected to reach $190 billion, a 61% increase from 2025, driven partly by rising memory costs, according to CNBC.
That $190 billion figure is worth sitting with for a moment.
For context, the entire gross domestic product of many mid-sized economies doesn’t reach that number in a year. Microsoft is committing that capital, predominantly to AI datacenter infrastructure, while Azure is already growing at 40%.
Here is a quick look at how the hyperscaler capital expenditure arms race has escalated:
- Microsoft: $190 billion projected for 2026, up 61% year over year, according to CNBC.
- Amazon: $200 billion estimated for 2026, unchanged from earlier forecast, per Bloomberg.
- Alphabet: $180-190 billion revised guidance for 2026, per, Bloomberg.
- Meta Platforms: $125-145 billion revised guidance for 2026, Business Insider highlighted.
- Combined Big Tech capex: Approaching $725 billion total in 2026, according to Bloomberg.
What the capex model actually implies for Azure
Here is where the Morgan Stanley analysis gets genuinely interesting for individual investors.
The firm ran what it calls a “capex-implied Azure AI revenue” analysis, essentially asking: given what Microsoft is spending on AI-specific infrastructure, and assuming various gross margin levels, what should Azure AI revenue look like by 2029?
The answer is uncomfortable for anyone who thinks current estimates already price in the AI opportunity.
At 40% gross margins on Azure AI, Morgan Stanley’s capex-implied revenue forecast for 2027 comes in 59% above their own bottom-up estimate, according to the firm’s report. At 50% margins, the implied upside reaches 91%.
I ran the numbers against the Morgan Stanley model myself and the key insight is this: the firm’s current Azure AI revenue estimate for fiscal year 2029 sits at roughly $117 billion. Their capex-implied figure, even at a conservative 20% gross margin, suggests $142 billion or more. The gap widens dramatically at higher margin assumptions.
Morgan Stanley maintained its Overweight rating and a $650 price target on Microsoft (MSFT), with shares trading around $416 as of late May 2026, according to the report. That implies roughly 56% upside from current levels if the thesis plays out.]
What Microsoft’s opportunity means for your portfolio
Microsoft stock has had a rough stretch. Shares are well below their 52-week high of $555.45, which means investors who have been cautious about AI spending have gotten some of what they wanted. A cheaper entry price.
But the Morgan Stanley framework essentially argues that caution is misplaced at this stage of the cycle.
Microsoft management has said it will “increase total AI capacity by over 80% this year, and roughly double our total datacenter footprint over the next two years,” according to the research report. That isn’t the language of a company hedging its AI bets.
The broader Wall Street consensus reflects similar confidence.
A survey of 31 analysts by public.com shows a Buy consensus and an average price target of $565.32 for MSFT, as of May 2026. Even Guggenheim, which flagged Q3 as a potential “mixed bag,” kept a Buy rating with a $586 target, according to TheStreet.
The point that most retail investors are missing is structural. Microsoft isn’t just selling cloud services anymore. It is building what amounts to a parallel electricity grid for artificial intelligence, and it controls a significant portion of the infrastructure that every AI application will eventually need to run on.
AI revenue for Microsoft, including Azure AI and Microsoft 365 Copilot, crossed a $37 billion annual run rate as of the most recent quarter. Six months ago, that figure barely registered as a line item.
The Morgan Stanley model projects that installed capacity could support revenues far beyond what current street estimates anticipate, so long as inference workloads scale and software attach across Copilot, Azure, Dynamics, and GitHub keeps improving.
That is a lot of “ifs.” But for a stock trading at a 40% discount to its recent high, the question isn’t whether Microsoft is a great business. The question is whether this level of infrastructure investment will ever get the revenue credit it deserves.
Morgan Stanley’s answer, backed by 46 pages of modeling, is yes. And the stock at $416 means the market hasn’t fully agreed yet.
